Template Argument Deduction

  Template Argument Deduction is the ability of templated classes to determine the type of the passed arguments for constructors without explicitly stating the type.

Before C++17, to construct an instance of a templated class we had to explicitly state the types of the argument (or use one of the make_xyz support functions).


std::pair p(2, 4.5);

Here, p is an instance of the class pair and is initialized with values of 2 and 4.5.


Or the other method of achieving this would be:


auto p = std::make_pair(2, 4.5);




Both methods have their drawbacks. Creating “make functions” like std::make_pair is confusing, artificial and inconsistent with how non-template classes are constructed. std::make_pair, std::make_tuple etc are available in the standard library, but for user-defined types it is worse: you have to write your own make_… functions.




In C++17, 


This requirement for specifying the types for a templated class constructor has been abolished. This means that we can now write:


auto p = std::pair(2, 4.5);



or


std::pair p(2, 4.5);


which is the logical way you would expect to be able to define p!




Example: 

So considering the earlier function mytuple(). Using Template Argument Deduction (and auto for function return type), consider:

Before C++ 17: 
auto mytuple()
{
   char a = 'a';
   int i = 123;
   bool b = true;
   return std::tuple(a, i, b);  // No types needed
}


C++ 17 :
This is a much cleaner way of coding – and in this case we could even wrap it as:


auto mytuple()
{
   return std::tuple(‘a’, 123, true);  // Auto type deduction from arguments
}